基于反应数据的贝叶斯概率方法在恢复力 模型选择中的应用

来源期刊:中南大学学报(自然科学版)2014年第5期

论文作者:刘佩 袁泉 魏庆朝

文章页码:1666 - 1672

关键词:贝叶斯理论;恢复力模型;模型选择;不确定性;最有可能模型

Key words:Bayesian theorem; restoring force model; model selection; uncertainty; the most probable models

摘    要:考虑模型选择过程中的不确定性,建立基于反应数据的模型选择的贝叶斯概率方法计算框架,用于选择一系列结构模型中的最有可能模型。该方法通过基于贝叶斯理论的模型参数识别方法得到模型参数的最有可能值及Hessian矩阵,对于全局可识别情况再结合渐近估计解法得到各模型的证据,进而通过贝叶斯定理得到选择各模型的概率,可以自动对过于复杂的模型进行限制。最后,将基于贝叶斯理论的模型选择方法用于采用实测滞回曲线数据的密肋复合墙试件的恢复力模型选择中,基于反应数据对选择2种恢复力模型的概率进行计算。研究结果表明:贝叶斯理论在系统识别中的应用不仅可以识别得到模型参数的最有可能值,还可以识别得到最有可能的模型,并且考虑了模型及模型参数的不确定性。

Abstract: Taking into account of uncertainties introduced in the model selection, the computational frame of Bayesian probability approach for model selection based on response data was proposed and used to obtain the most probable models from a set of models. Most probable values and Hessian matrix of model parameters were obtained through model parameter identification method based on Bayesian theorem. They were used to compute the evidence of a model through asymptotic approximation method for globally identifiable cases. The probability of a set of models was obtained according to the evidence. The proposed method automatically penalized a complicated model. The model selection method based on Bayesian theorem was applied to restoring force model selection of multi-grid composite wall specimen subjected to low cyclic loading with tested data. The probabilities of two restoring force models based on tested data were obtained. The results show that through application of Bayesian theorem to system identification, not only most probable values of model parameters can be obtained, but also the most probable models can be obtained. At the same time, uncertainties of model and model parameters can be considered.

有色金属在线官网  |   会议  |   在线投稿  |   购买纸书  |   科技图书馆

中南大学出版社 技术支持 版权声明   电话:0731-88830515 88830516   传真:0731-88710482   Email:administrator@cnnmol.com

互联网出版许可证:(署)网出证(京)字第342号   京ICP备17050991号-6      京公网安备11010802042557号